A Text Visualization Method For Cross-domain Research Topic Mining

Cross-domain research topic mining can help users find relationships among related research domains and obtain a quick overview of these domains.Some research domains have similar research topics with other domains,so some work published in a particular research domain may have influence on other research domains.The evolution of cross-domain topics are important for researchers to gain a deep understanding of their research domain,however no work focused on the cross-domain research topics analysis.This study focuses on the cross-domain research topic evolution;we design a visual analysis system and combine it with probabilistic topic model.A hierarchical topic model is adopted to extract topics of three different domains and to correlate the extracted topics.A simple yet effective visualization interface is then designed,which including word cloud,Sankey diagram,Treemap diagram and scatterplot,and certain interaction operations are provided to help users more deeply understand the visualization development trend and the correlation among the three domains.This study takes visualization,computer graphics and data mining as an example,using three case studies which include new topic finding,cross-domain evolution analysis and cross-domain topic analysis.The paper use visual analysis method to judge a research topic belongs to a specific domain or belongs to cross-domain.Users can use this system and some interactions with different views to explore some specific research topics,and gain the conclusion.The results of three case studies show that our method is effective.